Original subtitles

Next, I wanna take a look at the components that constitute a robot and

then try to analyze the effects of these components on the agility of the robot.

And more precisely, we'll be looking at the stopping distance of the robot.

So, this chart illustrates some of the design choices you might have if you go

online.

So I've chosen to visit dji.com and

pull off some frames that you can buy off the shelf.

DJI is currently the largest manufacturer of drones and

they have a wide selection of frames, batteries, propellers and motors.

In addition to that, you need an autopilot.

Pixhawk is one that's open source and sold by 3D Robotics.

You can also buy high level processors such as those made by Intel.

You clearly need the autopilot for doing low level control and you need something

like the Intel processor you see here to do high level computations.

You want to pay particular attention to the weight of each of these

because eventually your robot will have to carry these as it flies.

So the control architecture, you might think about,

involves using this low level processor to drive the motors and the propellers and

a high level processor like the Intel which communicates with

a lower level processor and commands a low level processor to drive the vehicle.

In addition you also want to have something

that complements the autonomous system.

A radio controller in case you have to take control of the vehicle.

So we show standard components here that you can buy off the shelf.

This is an example of an outdoor platform that we built.

And in this video you will essentially see that we have taken standard off the shelf

components, a DJI platform with motors that you can buy from the DJI website

with a simulated payload, it's a 600 gram payload and aluminum block

that simulates all the payload we might want to carry in the future.

It also has on board a 721 gram battery.

So this platform has a thrust to weight ratio which is greater than 2.7 and

this is important.

If you maximize the trust to weight ratio,

you essentially maximize the acceleration as we've seen before.

In addition to the processors, you also have to carry sensors.

Here we show two sets of sensors that we commonly use in our laboratory:

A laser scanner, and the laser scanner weighs about 270 grams.

A camera system that weighs about 80 grams.

And you also wanna think about power consumption and

there are two sources of power consumption.

First, the device itself consumes power.

It's roughly 10 watts for a laser scanner and 1.5 watts for the scanner system.

But in edition you're carrying these two payloads.

The fact that you're carrying a 270 gram laser scanner means you're burning

roughly 50 to 60 watts of power.

LIkewise the fact you're carrying an 80 gram camera means you're burning

roughly 15 watts of power.

So you wanna think about how heavy a sensor is, You also wanna see its range.

And all of those play into how fast the vehicle actually can go because

longer the range, longer the stopping distance can be.

You can detect obstacles far away, and therefore you have more time to come

to a stop if you see an obstacle in front of you.

This in turn allows you to go at a higher speed.

Of course, longer the range, the heavier your sensor might be, and

that in turn increases the weight of the platform.

While it increases the weight of the platform,

it'll decrease your trust to rate ratio.

So this in an interesting design space to explore.

So here are some examples of platforms we've built and

tested in our laboratory in the last year.

All these platforms are autonomous, they're different sizes,

they carry different sensors.

They weigh different amounts and they consume different amounts of power.